The columns
Each of the two seal columns is added only where it is absent, so a file made with them already skips the altering. The two identity facts are then written, whatever the file held before.
Nothing in the routine fills a seal in for a record that lacks one. So records written before the seals existed keep none, and the chain begins at the first record that has one.
What the next migration refuses
It makes the old deduplication table before it reads anything, so what follows reads a table that is certainly there.
Then two refusals. A legacy row with no matching fill refuses the whole migration, naming how many there were. A file spanning more than one symbol refuses too, naming how many.
Both refusals ask for the work to be done elsewhere rather than guessing. A row it cannot attribute and a file it cannot attribute to one symbol are both things a migration could paper over, and this one declines to.
What it records
The one symbol found. Where none was found at all, an empty one is recorded — so the fact is written either way, and what reads it later can tell nothing-found from not-yet-looked.
Related
Source evidence
---
row: P95-01
baseline: 49aa659
created: 2026-11-09 07:34 UTC
modified: 2026-11-09 07:34 UTC
evidence:
- claim: "Each of the two seal columns is added only where it is absent, so a file made with them already skips the altering"
source: "`event_spine.py` 586 into 587-588, and 589 into 590-591"
- claim: "The two identity facts are then written, whatever the file held before"
source: "`event_spine.py` 593-594, and 595-596"
- claim: "Nothing in the routine fills a seal in for a record that lacks one"
source: "`event_spine.py` 586-596, the whole of it"
- claim: "The later migration makes the old deduplication table before it reads anything"
source: "`event_spine.py` 621-622, ahead of 624"
- claim: "A legacy row with no matching fill refuses the whole migration, naming how many there were"
source: "`event_spine.py` 627 into 628-633"
- claim: "A file spanning more than one symbol refuses too, naming how many"
source: "`event_spine.py` 635 into 636-640"
- claim: "The one symbol found is recorded, and an empty one is recorded where none was found at all"
source: "`event_spine.py` 642, then 643-644"
---